How does
Pintrest
Use AI?
Increased product discovery and conversion by surfacing long-tail content.
Project Overview
Using multimodal gen AI to automatically generate contextual shopping collections aligned with how users search.
Layman's Explanation
Instead of relying on users to search using exact words, this AI system looks at all available products and content, understands them deeply using images and text, and creates curated shopping pages that match natural user queries like “summer wedding outfit in Italy.”
Details
PinLanding addresses the semantic mismatch between how users search (using contextual, occasion-driven phrases) and how e-commerce catalogs are organized (using static categories and tags). By using a combination of vision-language models (VLMs), CLIP, and large language models (LLMs), PinLanding interprets product images and descriptions and proactively creates “collections”—ready-to-serve landing pages centered around real-life search contexts. These AI-generated collections are scalable to millions of products and cost-efficient to run, enabling platforms to expose their long-tail inventory in meaningful, discoverable ways. This effectively shifts from reactive to proactive content discovery, bridging the intent gap in e-commerce and social platforms.
Analogy
It’s like a personal shopper at a department store who understands the style, season, and occasion you're dressing for, and instantly assembles racks of options for you—even before you ask.
Machine Learning Techniques Used
More Use Cases in
Technology
Novelty Justification
Multimodal GenAI for proactive, context-driven shopping is at the frontier of e-commerce innovation.
Project Estimates
